SNIP
The effective resolution of the 3200 flatbeds are more like
1600 dpi: less than 1/2 the resolution of the Nikon 8000/9000.
I've seen statements like this before, I've hardly ever seen any
proof.
Untill supported by evidence, I'll call them urban myths.
What does the resolution depend on?
1. It depends on captured resolution on film.
2. It depends on the scanner optics.
3. It depends on the CCD.
1. If the film doesn't have resolvable detail (lens+film and camera
shake). Maximum detail will be available only in the plane of focus,
and Depth-of-Field areas will not provide as high a resolution.
Because DoF depends on final output enlargement and viewing distance,
the best generalization one can make is by assuming e.g. general
viewing
conditions like normal reading distance (approx. 25cm or 10 inch) with
adequate object luminance.
2. Scanners use different technologies. Fixed focus, and scanning
through a glass plate, usually reduces image contrast which will also
cause fine detail (resolution) to suffer. Obviously, poor
lenses/prisms/mirrors will result in poor resolution. Also internal
reflection will reduce image contrast due to flare. Flare is often
poorly controlled in flatbed scanners.
3. Scanners use different sensor technologies. Some use so-called
staggered line sensor array placement, others have poor fill factors.
That will influence the MTF shape and aliasing tendency, while
limiting
resolution is mainly determined by the sampling density.
It is relatively simple to make an estimate about effective resolution
of the scanner itself, thus eliminating the influence of the film
resolution (which ultimately further reduces final resolution), by
scanning a sharp slanted edge and quantify the number of pixels it
needs to transition the edge (e.g. 10-90% brightness is a commonly
used criterion). Scan the very sharp edge at a small (approx. 5
degree) slant at the native scan resolution.
I tried several methods to produce such an edge target, folded
alumin(i)um foil is (too) vulnerable to keep flat in focus, so I
settled on a very sharp cutting blade, mounted in a slide mount. I
tried different blades, but found in a graphic arts shop very useful
"Logan" blades, model 270, in a 10-pack
(
http://www.logangraphic.com/products/blades/). Other blades as
offered in DIY shops may be equally useable, as long as they fit in a
slide mount which means they need to be thin and of the right size.
I happen to have, amongst others, an Epson 2450 (2400ppi) scanner,
which some claim to only have a *maximum* 1200ppi resolution (due to
the staggered sensor array). Scanning an edge as described above as a
transparency, produces the following profile:
<
http://www.xs4all.nl/~bvdwolf/temp/E2450_Edge.png> , so approx.
3.2-3.5 pixels wide, which is approaching 30 cy/mm or 62% of sensor
Nyquist *at that contrast* (which is impossible if it were a 1200ppi
sensor with a Nyquist frequency of 23.6, before the scanner lens
reduces that further).
Of course a more detailed analysis of the same edge is possible with a
program like Imatest (
www.imatest.com).
Bart
